A novel approach to calibrate a structured light vision sensor in a robot based 3D measurement system

A robot based 3D measurement system is presented to overcome the disadvantages of coordinate measuring machine (CMM). As to calibrate the vision sensor in the system, a novel procedure that utilizes the robot to generate sufficient calibration points with high accuracy is proposed. The approach is based on a number of fixed points printed in a paper which is of low-cost. A multilayer perceptron neural network (MLPNN) approach to calibrate the vision sensor is proposed. Various training algorithms for MLPNN are tested and the best performance one is employed to train the designed MLPNN. Using a standard part to validate the effectiveness of the presented technique, experimental results demonstrate designed MLPNN can achieve an accurate model for the structured light vision sensor. A robot based 3D measurement system is presented to overcome the disadvantages of coordinate measuring machine (CMM). As to calibrate the vision sensor in the system, a novel procedure that utilizes the robot to generate sufficient calibration points with high accuracy is proposed. The approach is based on a number of fixed points printed in a paper which is of low-cost. A multilayer perceptron neural network (MLPNN) approach to calibrate the vision sensor is proposed. Various training algorithms for MLPNN are tested and the best performance one is employed to train the designed MLPNN. Using a standard part to validate the effectiveness of the presented technique, experimental results demonstrate designed MLPNN can achieve an accurate model for the structured light vision sensor.